pyCarDisplay.detection package

Submodules

pyCarDisplay.detection.depth_detection_api module

This is the depth detection module.

We borrowed this code from the MiDaS GitHub repository, authored by intel-isl. We then modified the code to work with the pyCarDisplay library. This code obtains an image and turns it into a colorized heat map based on the depth of objects within the environment. Original functionality is preserved, but functionality is wrapped within a class with new additions.

Reference:

“MiDaS,” Pytorch.org. [Online]. Available: https://pytorch.org/hub/intelisl_midas_v2/. [Accessed: 27-Mar-2021].

Original license from MiDaS code is below.

MIT License

Copyright (c) 2019 Intel ISL (Intel Intelligent Systems Lab)

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

class pyCarDisplay.detection.depth_detection_api.DepthDetection(verbose: bool, model_path: str, model_type='large', optimize=True, model=None, device='cpu', transform=None, dpi=100, alpha=0.6, pixel_sizes=[1242, 375])[source]

Bases: object

Initialize the depth detection class.

verbosebool

Print out information regarding API activity.

model_pathstr

The path to the pretrained model file.

model_typestr, optional

The type of model to use to detect depth. The default is "large".

optimizebool, optional

Optimize the depth detection. The default is True.

modelNone, optional

If passed in, this is the predefined model. The default is None.

devicestr, optional

Use CUDA device if available. The default is "cpu".

transformNone, optional

Placeholder for the transform class variable. The default is None.

dpiint, optional

Dots per inch. The default is 100.

alphadouble, optional

The alpha variable. The default is 0.6.

pixel_sizeslist, optional

The sizes of the pixels. The default is [1242, 375].

None.

convert_to_heat_map(prediction, original_image)[source]
predictionImage

The depth predicitons of the objects withi the enviroment.

original_imageImage

The original image.

np.array()

A numpy array of the color values of each of the pixels in the image.

run(verbose: bool, pil_image: <module 'PIL.Image' from '/opt/anaconda3/envs/pyCarDisplay/lib/python3.8/site-packages/Pillow-8.1.2-py3.8-macosx-10.9-x86_64.egg/PIL/Image.py'>, optimize=True)[source]
verbosebool

Print out information regarding API activitty.

pil_imageImage

The image to be evaluated.

optimizebool, optional

Optimize the depth detection. The default is True.

predictionImage

An enhanced image with the colorized depth predictions of the objects within the environment.

pyCarDisplay.detection.object_detection_api module

This is the object detection module.

This code is borrowed from sgrvinod's GitHub repository named a-PyTorch-Tutorial-to-Object-Detection, and modified to work with pyCarDisplay. It preserves the original functionality, but is wrapped around a class.

Reference:

Vinodababu, S. (n.d.). A-PyTorch-Tutorial-to-Object-Detection. https://github.com/sgrvinod/a-PyTorch-Tutorial-to-Object-Detection

Original license from a-PyTorch-Tutorial-to-Object-Detection is below.

MIT License

Copyright (c) 2019 Sagar Vinodababu

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

class pyCarDisplay.detection.object_detection_api.ObjectDetection(model_path: str, verbose: bool, img_resize_size=(300, 300), norm_mean=[0.485, 0.456, 0.406], norm_std=[0.229, 0.224, 0.225], device='cpu')[source]

Bases: object

Initilize the object detection class.

model_pathstr

Path to the pre-trained PyTorch model.

verbosebool

Verbosity flag.

img_resize_sizetuple, optional

Image size during prediction. The default is (300, 300).

norm_meanlist, optional

Model hyper-parameter. The default is [0.485, 0.456, 0.406].

norm_stdlist, optional

Model hyper-parameter. The default is [0.229, 0.224, 0.225].

devicestr, optional

If 'gpu', model uses a CUDA device. The default is "cpu".

None.

detect(original_image: <module 'PIL.Image' from '/opt/anaconda3/envs/pyCarDisplay/lib/python3.8/site-packages/Pillow-8.1.2-py3.8-macosx-10.9-x86_64.egg/PIL/Image.py'>, suppress=None, min_score=0.2, max_overlap=0.5, top_k=200)[source]

Predicts the objects in the image using the pre-trained model.

original_imageImage

PIL image to be used for object detection.

suppressTYPE, optional

DESCRIPTION. The default is None.

min_scorefloat, optional

Minimum threshold for object to be classified. The default is 0.2.

max_overlapfloat, optional

Minimum threshold for object to be classified. The default is 0.5.

top_kint, optional

How many objects to attempt to classify. The default is 200.

dict

Returns a dictionary in the following format: {"annotated_image": annotated_image, "box_info": box_info, "detected": True} box_info is in the following format: box_info = {"text_size": text_sizes,"box_location": box_locations}, where the values are a list of coordinates for the detection boxes.

Module contents